WHAT THIS IS SAMPLE_SIGNAL | NOT_LIVE | NO_BACKTEST_VALIDATED
A 3D portfolio risk landscape. The risk surface is the PCA eigen-spectrum
(Ledoit-Wolf shrinkage + Marchenko-Pastur noise edge). The glowing tear-rings are
persistent-homology β1 cycles — topological "fractures" in the correlation network;
their count and intensity track the live fracture score f_t and its z-score.
The bars are HJB-Kelly rebalancing weights. All data is a
synthetic SAMPLE universe on a pure-Python CPU fallback — the cuML/cuPy/giotto-tda
GPU path is ROADMAP. This is NOT live trading and is
NOT backtest-validated. Λ = Conjecture 1 (< 1.0); trust < 100%.
loading live quant signal…
PCA RISK Ledoit-Wolf · MP
…
TDA FRACTURE β0/β1 persistence
…
HJB-KELLY w* sizing
ENERGY / SIGNED RECEIPT DSSE-style
…
COMPUTE BACKEND CPU fallback
…
Build references (no runtime calls): RAPIDS cuDF/cuML
(docs.rapids.ai),
giotto-tda (giotto-ai/giotto-tda),
Ledoit & Wolf (2003) shrinkage covariance, Gidea & Katz (2017) TDA of crashes (arXiv:1703.04385),
Marchenko-Pastur noise edge. β1 here is the 1-skeleton cycle rank (honest fast proxy);
the exact Vietoris-Rips H1 is the GPU ROADMAP path.